[config] math_info test issue
Hi, I'm looking at libs/config/test/Jamfile, and it starts with this: # *** DO NOT EDIT THIS FILE BY HAND *** However, looking at the libs/config/tools/generate.cpp I don't see any "math_info" test being generated. Further, changelog contains this: 2005-12-06 20:01 johnmaddock * test/Jamfile: Added new file math_info to probe for known or suspected math-related problems and setup. There's no associated change to generate.cpp. See seems like the file was edited by hand. Am I wrong? - Volodya
There's no associated change to generate.cpp. See seems like the file was edited by hand. Am I wrong?
Oh !****?:@! My mess: yes that file is generated programmatically, and as the author of the said program I should have known better. I guess we need to modify the generator program to produce a bbv2 file? John.
John Maddock wrote:
There's no associated change to generate.cpp. See seems like the file was edited by hand. Am I wrong?
Oh !****?:@!
My mess: yes that file is generated programmatically, and as the author of the said program I should have known better. I guess we need to modify the generator program to produce a bbv2 file?
Yes, but I'll take care of that. I've already started but was stopped short by absence of math_info in generate.cpp, and though that maybe generate.cpp is now obsoleted and that I'd better ask. Thanks, Volodya
Yes, but I'll take care of that. I've already started but was stopped short by absence of math_info in generate.cpp, and though that maybe generate.cpp is now obsoleted and that I'd better ask.
It should be a trivial change: just alter the boilerplate at the start of write_jamfile, I'm assuming that the remaining test cases should all be the same in bbv2? John.
John Maddock wrote:
Yes, but I'll take care of that. I've already started but was stopped short by absence of math_info in generate.cpp, and though that maybe generate.cpp is now obsoleted and that I'd better ask.
It should be a trivial change:
Yes.
just alter the boilerplate at the start of write_jamfile, I'm assuming that the remaining test cases should all be the same in bbv2?
Not quite. V2 does not have <template> but has project-level requirements, so that should be changed as well. Trivial change too, already done and comitted. But: when I run the program, the generated Jamfile (*not* Jamfile.v2), is very different from the existing one. Maybe, you can run the program, check if the differences are fine, and commit them? After that I'll take care of Jamfile.v2. There's another problem, in the generated Jamfile there are two test suites called HAS_TR1_ARRAY and V2 is less forgiving. Maybe, the test suite names should be all unique? - Volodya
But: when I run the program, the generated Jamfile (*not* Jamfile.v2), is very different from the existing one. Maybe, you can run the program, check if the differences are fine, and commit them? After that I'll take care of Jamfile.v2.
Done, and committed: when I ran the *only* differences in the V1 Jamfile were the datestamps, and that's the way it should be.
There's another problem, in the generated Jamfile there are two test suites called
HAS_TR1_ARRAY
and V2 is less forgiving. Maybe, the test suite names should be all unique?
It's a SNAFU in a test case, fixed now. John.
John Maddock wrote:
But: when I run the program, the generated Jamfile (*not* Jamfile.v2), is very different from the existing one. Maybe, you can run the program, check if the differences are fine, and commit them? After that I'll take care of Jamfile.v2.
Done, and committed: when I ran the *only* differences in the V1 Jamfile were the datestamps, and that's the way it should be.
Thanks. The reason why Jamfile was altered when I run it turned out to be very simple -- the order in which the program find test files depends on the order the filesystem lists them. - Volodya
participants (2)
-
John Maddock -
Vladimir Prus